DNN Task Assignment in UAV Networks: A Generative AI Enhanced Multi-Agent Reinforcement Learning Approach

TL;DR

This work addresses the challenge of low-latency, energy-aware DNN inference in UAV networks by proposing a two-stage framework that decouples path planning from task assignment. The path planning stage uses a task-size-aware greedy algorithm to minimize flight distance, while the second stage introduces GDM-MADDPG, which replaces the MADDPG actor with a reverse denoising process from a generative diffusion model to generate DNN task assignments in a dynamic UAV environment. Results from simulations show that GDM-MADDPG improves AoI, task completion rates, energy efficiency, and load balancing compared to baselines, validating the effectiveness of integrating diffusion-based decision generation with multi-agent reinforcement learning. The approach enables scalable, robust collaborative DNN inference in UAV swarms with HAP-assisted edge computing, with practical impact for real-time sensing, disaster response, and IoT applications.

Abstract

Unmanned Aerial Vehicles (UAVs) possess high mobility and flexible deployment capabilities, prompting the development of UAVs for various application scenarios within the Internet of Things (IoT). The unique capabilities of UAVs give rise to increasingly critical and complex tasks in uncertain and potentially harsh environments. The substantial amount of data generated from these applications necessitates processing and analysis through deep neural networks (DNNs). However, UAVs encounter challenges due to their limited computing resources when managing DNN models. This paper presents a joint approach that combines multiple-agent reinforcement learning (MARL) and generative diffusion models (GDM) for assigning DNN tasks to a UAV swarm, aimed at reducing latency from task capture to result output. To address these challenges, we first consider the task size of the target area to be inspected and the shortest flying path as optimization constraints, employing a greedy algorithm to resolve the subproblem with a focus on minimizing the UAV's flying path and the overall system cost. In the second stage, we introduce a novel DNN task assignment algorithm, termed GDM-MADDPG, which utilizes the reverse denoising process of GDM to replace the actor network in multi-agent deep deterministic policy gradient (MADDPG). This approach generates specific DNN task assignment actions based on agents' observations in a dynamic environment. Simulation results indicate that our algorithm performs favorably compared to benchmarks in terms of path planning, Age of Information (AoI), energy consumption, and task load balancing.

Figures (8)

• Figure 1: Four task assignment models.
• Figure 2: The computational complexity, processing latency, and output data size of each layer of different models.
• Figure 3: Network architecture.
• Figure 4: GDM-MADDPG for strategy generation.
• Figure 5: UAV path planning considering task size and flight distance.